A WEAKLY-SUPERVISED DEEP NETWORK FOR DSM-AIDED VEHICLE DETECTION
DSM辅助车辆检测的弱监督深层网络
论文:http://static.tongtianta.site/paper_pdf/003c66ae-eb5f-11e9-9dd0-00163e08bb86.pdf
ABSTRACT
摘要
With the breakthrough of the spatial resolution of optical remote sensing images at the sub-meter level and the explosive development of deep learning, geospatial object detection has achieved a growing interest in remote sensing community. However, labeling large training datasets in object level is still an expensive and tedious procedure. This might lead to the poor model generalization and degraded network learning ability. To this end, a weakly-supervised deep network (WSDN) is developed for geospatial object detection by applying a digital surface model (DSM)-aided auto-labeling and a pre-trained network learned from the task-independent dataset. Experimental results conducted on the stereo aerial imagery of a large camping site are performed to demonstrate that the proposed WSDN yields better detection results, with 62.78% precision and 55.13% recall.
随着亚米级光学遥感图像空间分辨率的突破以及深度学习的爆炸性发展,地理空间物体检测在遥感界引起了越来越多的兴趣。但是,在对象级别标记大型训练数据集仍然是昂贵且繁琐的过程。这可能导致较差的模型泛化和降低的网络学习能力。为此,通过应用数字表面模型(DSM)辅助的自动标记和从独立于任务的数据集中学习的预训练网络,开发了一种用于地理空间物体检测的弱监督深度网络(WSDN)。对大型露营地的立体航拍图像进行的实验结果表明,所提出的WSDN具有更好的检测结果,精度为62.78%,召回率为55.13%。
Index Terms— Deep learning, digital surface model, geospatial object detection, optical remote sensing imagery, vehicle, weakly-supervised
索引词-深度学习,数字表面模型,地理空间物体检测,光学遥感影像,车辆,弱监督
1. INTRODUCTION
1.引言
Recently, optical remote sensing imagery (RSI) has been paid a growing interest in many applications, such as urban mapping and monitoring [1], mineral exploration [2, 3], particularly spatial object detection [4]. Existing detection methods can be roughly categorized as follows [5]: template matchingbased,knowledge-based, object-based, and machine learningbased methods. The explosive development in deep learning have made them unsurprisingly applied to object detection in RSIs [6] and have shown a stronger detection performance than the aforementioned traditional methods. Labeling training datasets [7, 8] plays an important role in object detection in optical RSIs. In addition, for objects in optical RSIs with a cluttered background, relatively small ground sampling distance (GSD), and various deformations, e.g. variabilities in viewpoint, scaling, and direction, their labeling problems have always been challenging and existing manual labeling is not only time-consuming and laborious but also inconsistent
最近,光学遥感影像(RSI)在许多应用中引起了越来越多的兴趣,例如城市制图和监视[1],矿物勘探[2,3],尤其是空间物体检测[4]。现有的检测方法可以大致分类如下[5]:基于模板匹配,基于知识,基于对象和基于机器学习的方法。深度学习的爆炸性发展使得它们毫无疑问地应用于RSI中的对象检测[6],并且比上述传统方法具有更强的检测性能。标记训练数据集[7、8]在光学RSI中的对象检测中起着重要作用。此外,对于背景杂乱的光学RSI中的对象,相对较小的地面采样距离(GSD),以及各种变形,例如视点,缩放比例和方向的可变性,它们的标签问题一直是具有挑战性的,并且现有的手动标签不仅费时费力,而且不一致
所有论文

添加客服微信,加入用户群
蜀ICP备18016327号